基于学习的霍夫变换线段组物体检测算法

An Object Detection Algorithm of Hough Transform Line Segmentation Groups Based on Learning

针对单条霍夫变换线段特征算法的区分能力弱,不能有效处理部分匹配等问题,提出了霍夫变换线段组算法。首先通过文中算法提取霍夫变换线段特征构成码表,以此码表作为弱检测器的输入,再通过AdaBoost算法学习将弱检测器构造成强检测器,以提高检测的效率,最后在测试集上进行检测。为了计算两条霍夫变换线段之间的相似度,引入四元组空间内加权欧式距离,通过合理调整权重,能够有效地处理不可靠边缘检测问题。实验表明该算法能处理部分遮挡问题,具有很好的发展潜力。

Aiming at the problems of the weak distinguishing ability for the algorithm based on single Hough Transform Line Segment （HTLS） feature, which cannot effectively deal with partial matching, an algorithm of the HTLS groups is proposed. Firstly in this paper, the algorithm extracts the Hough transform line segment feature to constitute the codebook as input of weak detector. Then through the study of AdaBoost algorithm make weak detectors structure into a strong detector,in order to improve the efficiency of detection. The fi- nal tests on the test set. To calculate the similarity between the two Hough transform line segment, a weighted Euclidean distance is intro- duced, through adjusting the weights, can effectively deal with unreliable edge detection problem. The experiment shows that the algorithm can deal with the partial sheltering problem, has a very good development potential.